Light emitting diodes (LEDs) based on crystalline inorganic semiconductors provide numerous advantages, including high brightness, long lifetime, and high efficiency. This makes inorganic LEDs good candidates for display and lighting applications. Unfortunately, inorganic LEDs generally emit light at only one wavelength (i.e., a particular color), so using them for white light solid state lighting or polychromatic LED displays usually involves integration of multiple LEDs, each of which is configured to emit light at a distinct wavelength. For example, each pixel in an existing color LED display usually has three LED elements (or subpixels) to emit red, green, and blue light, respectively. These subpixels are aligned in next to each other, e.g., in a Bayer array. However, this parallel configuration increases the size of each pixel, thereby limiting the spatial resolution of the resulting display. In addition, since light from these subpixels is separate near the display surface and then mixed in the far field, the color accuracy of the display is also compromised.